Forklift Pedestrian Safety Technology Compared: AI Video, Proximity Tags, Blue Lights and Radar

Four technologies dominate forklift-pedestrian safety: blue or red warning lights, proximity tags, on-truck radar or AI cameras, and fixed AI video analytics. They differ in what they detect, who they warn and how fast. Lights only signal, tags and on-truck sensors warn at the moment of risk, and fixed video mostly produces data for supervisors. None of them replaces physical separation of people and trucks.

Why forklift-pedestrian collisions get special attention

Forklifts cause fewer injuries than manual handling or slips, but the injuries are severe. A Bureau of Labor Statistics (BLS) fact sheet counted 614 US workers killed in forklift-related incidents from 2011 to 2017, including 74 deaths in 2017 ^[1]. In 2017, forklifts were involved in 9,050 nonfatal injuries with days away from work, and 1,850 of those were pedestrian incidents ^[1]. Pedestrians injured by forklifts had a median of 20 days away from work, compared with 13 days for all forklift cases and 8 days for all workplace injuries ^[1]. BLS changed its injury classification starting with 2023 data, so newer figures are not directly comparable with these.

OSHA lists being struck by a lift truck among the ways workers are injured around these machines, along with trucks driven off loading docks and falls from elevated pallets and tines ^[3]. The underlying problem is visibility. A loaded counterbalance truck blocks much of the driver's forward view, racking ends create blind corners, and pedestrians often assume a driver has seen them.

What do the rules require?

In the US, the powered industrial truck standard, 29 CFR 1910.178, puts the duty on the driver and the employer's traffic rules. Drivers must "slow down and sound the horn at cross aisles and other locations where vision is obstructed" and must look in the direction of travel and keep a clear view of the path ^[2]. Trucks must not be driven up to anyone standing in front of a bench or other fixed object, and each operator's performance must be evaluated at least once every three years ^[2]. The standard does not require warning lights, proximity systems or cameras.

In Great Britain, HSE workplace transport guidance says roadways and footpaths should be separate whenever possible, and that the most effective approach is to make pedestrian and vehicle routes entirely separate ^[4]. It lists protective barriers, clear markings and raised kerbs as ways to keep vehicles out of pedestrian areas, and suggests traffic lights, marked crossings, bridges or subways at busy crossing points ^[4]. When specifying vehicles, HSE asks employers to consider which warning systems, such as horns and lights, are fitted ^[6].

So the technologies in this article are additional layers. They matter most where separation is incomplete: shared aisles, dock areas, doorways and crossing points.

How do blue and red warning lights work?

Warning lights project a colored spot, arrow or line onto the floor around a moving truck. Linde Material Handling, for example, describes BlueSpot as an LED mounted on the overhead guard that projects a blue (or red) dot onto the ground to warn that a truck is approaching, and sells red warning lines that project safety lines along the sides of the truck so pedestrians can see how much distance to keep ^[7].

A warning light detects nothing. It does not know whether a person is nearby, and it gives the driver no information. Its value depends on a pedestrian looking at the floor, understanding the signal and reacting. That makes it closer to a sign or a floor marking than to a detection system. It can help at blind corners and rack ends, where the light appears before the truck does.

The limits are practical. A light that is on all the time on every truck can become background, much as HSE notes that reversing alarms "may be so common on a busy site that pedestrians do not take any notice" ^[5]. Bright ambient light and dark or reflective floors reduce contrast. Manufacturer pages describe what the lights do, but Linde's light and warning page gives no effectiveness figures or detection distances ^[7]. Lights are cheap and simple to fit, which is a fair reason to use them, provided nobody counts them as a control that detects people.

How do proximity tags work?

Proximity warning systems put a tag on each person and a unit on each truck, then measure the distance between them. Most current forklift systems use ultra-wideband (UWB) radio, which measures time of flight between devices and gives accurate ranging. Modjoul says its HaloGuard system uses UWB in the 6240 to 6739.2 MHz range, with tags worn on its SmartBelt, modules on forklifts, and alerts to both drivers and pedestrians through vibration, sound and light ^[9]. Modjoul also says HaloGuard offers automatic speed control for vehicles and location tracking to find hotspots for worker-forklift interactions ^[9].

Tags have two strengths. The distance measurement is precise, and the warning reaches both people at the moment of risk, which few other technologies do. The interaction logs also show where near approaches cluster, which helps with traffic planning.

The weakness is coverage. A tag system protects only people who are wearing a charged tag. Visitors, drivers from other companies, contractors and anyone who left their tag in a locker are invisible to it. Sites using tags need a process for issuing tags at entry, checking battery charge between shifts, and handling people who arrive without one. Wearing a tag also creates location data tied to a named person, which brings privacy duties discussed in the guide's chapter on privacy and workers.

How do radar and AI cameras on the truck work?

On-truck sensors detect objects without anyone wearing a tag.

Radar sends out radio waves and measures reflections. It sees through dust and darkness, but it detects every object in its field: racking, pallets, walls and parked trucks as well as people. HSE advises that radar is useful as a reversing aid "on open sites where the number of unwanted alarms is likely to be low" ^[5]. That caveat has a long research history. In a two-year NIOSH evaluation of a radar proximity warning system on large mining trucks, published in 2006, the system detected people and vehicles reliably, but alarms from objects that posed no danger were common, and the author recommended pairing radar with cameras so operators could check the cause of an alarm ^[10]. In a dense warehouse aisle, the same problem is likely to be worse. Linde describes a Reverse Assist Radar that monitors the area behind the truck, warns the driver and can brake in an emergency ^[8].

AI cameras on the truck use object detection to identify people specifically. Linde describes a Reverse Assist Camera that automatically detects people when the truck reverses, warns the driver and can be set to reduce speed, with a front-facing version as well ^[8]. OneTrack sells AI cameras and sensors mounted on forklifts, with an emphasis on operational analytics alongside safety ^[13]. Because they recognize people, these systems should produce fewer nuisance alarms than radar. They inherit the usual camera limits. HSE notes for vehicle CCTV that a dirty lens makes a camera much less effective, that systems may need time to adjust when moving from dark to bright areas, and that drivers can find heights and distances hard to judge on screen ^[5].

Both radar and cameras usually warn only the driver. The pedestrian learns nothing unless the truck sounds a horn or stops.

How does fixed AI video analytics work?

Fixed AI video analytics runs computer vision on ceiling or wall cameras, often the site's existing CCTV. Protex AI says it integrates with existing CCTV in logistics facilities and lets users define safety zones for people and vehicles, with real-time alerts when limits are breached ^[11]. Typical events include forklifts entering pedestrian walkways, failure to stop at intersections, speeding and people walking outside marked routes.

In practice, most alerts go to supervisors or a dashboard. The driver and pedestrian are not warned in the moment unless the system drives a light or sound at the location. The main output is data: counts of near misses by place, time and type, with video clips that show what happened.

Published results come from vendors and should be read as claims. Voxel reports that at a Verst Logistics warehouse in Hebron, Kentucky, daily vehicle safety incidents (detected events, not injuries) fell from 19 to 3 in six months, with an 82% drop in vehicle safety incidents and a 92% drop in "no stop at intersection" events ^[12]. Voxel attributes the change to video-based coaching of forklift drivers, more stop signs, driver retraining and marked no-pallet zones to keep pedestrian paths clear ^[12]. Protex AI attributes an 80% reduction in incidents to Marks & Spencer and reports large drops in near-miss events for unnamed customers ^[11]. These figures have no control group and are not independently verified. The Verst example does show the usual mechanism: the improvement came from the layout and training changes that managers made after seeing the data.

How do the four technologies compare?

Which combination fits which site?

The right mix depends on how much pedestrian and truck traffic overlaps and who the pedestrians are.

• Sites with good physical separation and few crossing points often get most value from better barriers and crossing design, with warning lights as a low-cost aid at blind corners.
• Sites where a stable, trained workforce shares aisles with trucks, such as pick areas, are where tags fit best, because everyone can be issued a tag and the system can warn both parties.
• Sites with many visitors, agency staff or truck drivers on foot, such as docks and yards, need detection that does not depend on tags: on-truck cameras or radar, or fixed analytics at the crossings.
• Multi-site operators trying to find where risk concentrates often start with fixed video analytics on existing cameras, then use the data to decide where engineering changes or on-truck systems are worth funding.

Combining layers is common. A tag or camera system handles the moment of risk, and fixed analytics shows whether the layout and rules are working.

What to check before buying

Ask each vendor for detection results under your conditions: your lighting, your floor, your truck types and your aisle widths. Ask for false alarm rates per hour of driving, not only detection rates, because drivers stop trusting systems that alarm too often. Ask how the system behaves when a tag is missing, a lens is dirty or a sensor fails, and whether failures are reported.

For any system that slows or stops a truck, involve the truck manufacturer and an engineer. Automatic braking with a raised load or on a ramp creates hazards of its own. In the EU, Regulation (EU) 2023/1230 on machinery replaces the Machinery Directive and applies from 20 January 2027; under the current Directive, machinery modified to the extent that it becomes, in effect, new machinery falls back into scope ^[14]. Retrofits that change how a truck moves should be checked against those rules.

Finally, agree in writing how the data will be used. Tags and cameras both create records of individual behavior. Workers are more likely to wear tags and accept cameras when the data is used for layout changes and coaching, and less likely when it is used mainly for discipline.

Summary

Warning lights signal but do not detect. Proximity tags measure distance precisely and warn both driver and pedestrian, but only for people wearing charged tags. On-truck radar needs no tags but alarms on everything, while on-truck AI cameras recognize people and depend on lighting and clean lenses. Fixed AI video analytics covers everyone in view and produces the best data for redesigning traffic routes, though it rarely warns anyone in the moment. US and UK rules put physical separation and driver behavior first, and published performance figures for most of these products come from vendors.